CN-115460546-B - Urban mobile network traffic generation method and device, electronic equipment and storage medium

CN115460546BCN 115460546 BCN115460546 BCN 115460546BCN-115460546-B

Abstract

The invention provides a method, a device, electronic equipment and a storage medium for generating urban mobile network traffic, wherein the method comprises the steps of extracting a plurality of entities based on urban data information and constructing an urban knowledge graph based on the relationship among the entities; the method comprises the steps of obtaining a knowledge spectrum embedded vector corresponding to a base station based on a city knowledge spectrum, inputting the knowledge spectrum embedded vector and noise into a trained generator network for generating an countermeasure network model to obtain generated flow data, generating a countermeasure network and a continuous time city mobile network flow generation countermeasure network based on the knowledge spectrum embedded vector of the base station transmitting known network flows in different time scales in the countermeasure network model, inputting the knowledge spectrum embedded vector into the generator network to obtain generated flow sample data, judging real flow data and the generated flow sample data based on a discriminator network, and performing countermeasure training. The invention can generate reliable urban mobile network traffic.

Inventors

LI YONG
HE YUEYING
LIU ZHONGJIN
Hui Lidi
WANG HUANDONG
JIN DEPENG
ZHANG JIAQI
ZHANG JIANSONG
Xing Yanzhen
ZOU ZHE

Assignees

清华大学
国家计算机网络与信息安全管理中心

Dates

Publication Date: 20260508
Application Date: 20220719

Claims (12)

1. A method for generating urban mobile network traffic, the method comprising: extracting a plurality of entities based on urban data information, and constructing an urban knowledge graph based on relationships among the entities, wherein the entities at least comprise a base station, an area, a business district, POIs, POI categories and brands; obtaining a knowledge graph embedding vector corresponding to the base station based on the urban knowledge graph; The knowledge graph embedded vector and noise are input to a trained generator network for generating an countermeasure network model, and generated flow data output by the generator network is obtained, wherein the generated flow data is urban mobile network flow under the urban data information background; The system comprises a generating countermeasure network model, a continuous time city mobile network traffic generating countermeasure network and a communication network, wherein the generating countermeasure network model comprises city mobile network traffic pattern generating countermeasure network of different time scales and continuous time city mobile network traffic generating countermeasure network, the city mobile network traffic generating countermeasure network is based on knowledge graph embedded vectors of a plurality of base stations for transmitting known network traffic, the generating traffic sample data is obtained, real traffic data and the generating traffic sample data are subjected to countermeasure training based on a discriminator network, the real traffic data are traffic data transmitted by the base stations for transmitting known network traffic, the city mobile network traffic pattern generating countermeasure network of different time scales comprises city mobile network traffic pattern generating countermeasure network of day scales and city mobile network traffic pattern generating countermeasure network of week scales, the continuous time city mobile network traffic generating countermeasure network refers to city mobile network traffic generating countermeasure network under continuous time, wherein the time length of the continuous time is longer than the week; the decision-based network performs countermeasure training on the real flow data and the generated flow sample data, and specifically includes: Carrying out projection processing on the real flow data and the generated flow sample data to respectively obtain a real flow data projection vector and a generated flow sample data projection vector; processing the real flow data projection vector and the generated flow sample data projection vector through a multi-layer perceptron, layer normalization and an activation function to obtain a discrimination result output by the discriminator network; And generating an countermeasure network for the urban mobile network traffic patterns with different time scales based on the discrimination result and the loss function, wherein the Wasserstein distance added with gradient penalty is adopted as the loss function.
2. The method for generating urban mobile network traffic according to claim 1, wherein the obtaining the knowledge-graph embedded vector corresponding to the base station based on the urban knowledge-graph specifically comprises: and obtaining a knowledge spectrum embedding vector corresponding to the base station through a knowledge spectrum embedding model based on the urban knowledge spectrum.
3. The urban mobile network traffic generation method according to claim 1, wherein said loss function is determined using the following formula: ; Wherein, the Representing the loss function; representing a discrimination result of the real flow data; representing the discrimination result of the generated flow sample data; Representing a distribution of the real flow data; representing a distribution of the generated traffic sample data; representing uniform sampling between true-generated sample pairs; Representing the discrimination result of the sample; representing the distribution of samples; Representing penalty term weights; representing a gradient operation; Representing a desired operation; representing a two-norm operation.
4. The method of claim 1, wherein the generating the countermeasure network for the urban mobile network traffic patterns of different time scales comprises generating the countermeasure network for the urban mobile network traffic patterns of the day scale, wherein the traffic patterns of the base stations in the plurality of days are averaged in the time scales of each day for the urban mobile network traffic patterns of the day scale, wherein the generator network comprises a first generator network, wherein the discriminator network comprises a first discriminator network, wherein the generated traffic sample data comprises urban mobile network traffic pattern sample data of the day scale, wherein the discrimination results comprise a first discrimination result; The knowledge graph embedded vector of the base station based on the plurality of transmission known network flows is input into a generator network to obtain generated flow sample data, and the method specifically comprises the following steps: Presetting a first preset flow mode matrix, wherein the first preset flow mode matrix is an initialized flow mode taking a day as a scale; the knowledge spectrum embedded vector of the base station transmitting the known network flow and noise are input into the first generator network, and the knowledge spectrum embedded vector of the base station transmitting the known network flow and the noise are converted into a first projection vector through a multi-layer perceptron, layer normalization and activation function; multiplying the first projection vector and the first preset flow pattern matrix, and then performing activation function processing to obtain urban mobile network flow pattern sample data taking the day as a scale; the decision-based network performs countermeasure training on the real traffic data and the generated traffic sample data, and specifically includes: Carrying out projection processing on the real flow data and the urban mobile network flow mode sample data with the day as a scale to respectively obtain a real flow data projection vector and an urban mobile network flow mode sample data projection vector with the day as a scale; Processing the real flow data projection vector and the urban mobile network flow mode sample data projection vector taking the day as a scale by a multi-layer perceptron, layer normalization and activation functions to obtain a first discrimination result output by the first discriminator network; And generating an countermeasure network for countermeasure training for the urban mobile network traffic mode with the scale of days based on the first discrimination result and the loss function.
5. The urban mobile network traffic generation method according to claim 4, wherein said first predetermined traffic pattern matrix is determined by: clustering the real flow data based on a clustering algorithm to obtain a plurality of first clustering centers; and taking the plurality of first clustering centers as initial values of each basis vector in the first preset flow pattern matrix, and obtaining the first preset flow pattern matrix based on the initial values.
6. The method of claim 4, wherein the generating the countermeasure network for the urban mobile network traffic patterns of different time scales comprises generating the countermeasure network for the urban mobile network traffic patterns of a week scale, wherein the traffic transmitted by the base station in the plurality of weeks for the urban mobile network traffic patterns of the week scale is averaged for the time scales of the week, wherein the generator network comprises a second generator network, wherein the discriminator network comprises a second discriminator network, wherein the generating the traffic sample data comprises the urban mobile network traffic pattern sample data of the week scale, and wherein the discrimination result comprises a second discrimination result; The knowledge graph embedded vector of the base station based on the plurality of transmission known network flows is input into a generator network to obtain generated flow sample data, and the method specifically comprises the following steps: presetting a second preset flow mode matrix, wherein the second preset flow mode matrix is an initialized flow mode which does not contain the flow mode with the day as a scale and takes the week as a scale; Inputting the knowledge spectrum embedded vector of the base station transmitting the known network traffic and noise into the second generator network, and converting the knowledge spectrum embedded vector of the base station transmitting the known network traffic and the noise into a second projection vector through a multi-layer perceptron, layer normalization and activation function; Multiplying the second projection vector and the second preset flow pattern matrix, and then performing activation function processing to obtain week-scale flow pattern sample data without the day-scale flow pattern; obtaining the week-scale urban mobile network traffic pattern sample data based on the repeated sample data of the week-scale urban mobile network traffic pattern sample data without the day-scale traffic pattern; the decision-based network performs countermeasure training on the real traffic data and the generated traffic sample data, and specifically includes: Carrying out projection processing on the real flow data and the urban mobile network flow mode sample data taking the week as a scale to respectively obtain a real flow data projection vector and an urban mobile network flow mode sample data projection vector taking the week as a scale; Processing the real flow data projection vector and the city mobile network flow mode sample data projection vector taking the circumference as a scale through a multi-layer perceptron, layer normalization and an activation function to obtain a second discrimination result output by the second discriminator network; And generating an countermeasure network for countermeasure training for the city mobile network traffic mode with the week scale based on the second discrimination result and the loss function.
7. The urban mobile network traffic generation method according to claim 6, wherein said second predetermined traffic pattern matrix is determined by: clustering the real flow data based on a clustering algorithm to obtain a plurality of second aggregation centers; and taking the plurality of second aggregation centers as initial values of each base vector in the second preset flow mode matrix, and obtaining the second preset flow mode matrix based on the initial values.
8. The urban mobile network traffic generation method according to claim 6, wherein the generator network comprises a third generator network, wherein the discriminator network comprises a third discriminator network, wherein the generated traffic sample data comprises continuous-time urban mobile network traffic sample data, wherein the discrimination results comprise a third discrimination result; The knowledge graph embedded vector of the base station based on the plurality of transmission known network flows is input into a generator network to obtain generated flow sample data, and the method specifically comprises the following steps: The knowledge graph embedded vector and noise of the base station transmitting the known network flow are input to the third generator network, and a converted sequence is obtained through a multi-layer perceptron and layer normalization, wherein the sequence length of the converted sequence is the same as the sequence length of the continuous-time urban mobile network flow sample data; The converted sequence is used as an initial value to be input into three time convolution neural networks with different convolution kernel sizes, and the time convolution neural networks are processed through a multi-layer perceptron and a first activation function to obtain continuous time urban mobile network flow fluctuation sample data; Obtaining continuous-time urban mobile network flow sample data based on the continuous-time urban mobile network flow fluctuation sample data and repeated sample data of the urban mobile network flow mode sample data taking the week as a scale; the decision-based network performs countermeasure training on the real traffic data and the generated traffic sample data, and specifically includes: Inputting the real flow data and the continuous-time urban mobile network flow sample data into three time convolution neural networks with different convolution kernel sizes, and processing the three time convolution neural networks through a multi-layer perceptron and a second activation function to obtain a third discrimination result output by the third discriminator network; and generating an countermeasure network for countermeasure training for the continuous-time urban mobile network traffic based on the third discrimination result and the loss function.
9. An urban mobile network traffic generating device, characterized in that it is configured to implement the urban mobile network traffic generating method according to any one of claims 1 to 8, said device comprising: The building module is used for extracting a plurality of entities based on the urban data information and building an urban knowledge graph based on the relationship among the entities, wherein the entities at least comprise a base station, an area, a business circle, POIs, POI categories and brands; the processing module is used for obtaining a knowledge graph embedding vector corresponding to the base station based on the urban knowledge graph; the training module is used for inputting the knowledge graph embedded vector and noise into a trained generator network for generating an countermeasure network model to obtain generated flow data output by the generator network, wherein the generated flow data is urban mobile network flow under the urban data information background; the generating countermeasure network model comprises a city mobile network traffic pattern generating countermeasure network with different time scales and a continuous time city mobile network traffic generating countermeasure network, wherein the city mobile network traffic pattern generating countermeasure network with different time scales and the continuous time city mobile network traffic generating countermeasure network are both input into a generator network based on knowledge graph embedded vectors of a plurality of base stations transmitting known network traffic to obtain generated traffic sample data, and real traffic data and the generated traffic sample data are subjected to countermeasure training based on a discriminator network to obtain the real traffic data, wherein the real traffic data are traffic data transmitted by the base stations transmitting the known network traffic.
10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the urban mobile network traffic generation method according to any one of claims 1 to 8 when the program is executed by the processor.
11. A non-transitory computer readable storage medium having stored thereon a computer program, which when executed by a processor implements the urban mobile network traffic generation method according to any of claims 1 to 8.
12. A computer program product comprising a computer program which, when executed by a processor, implements the urban mobile network traffic generation method according to any one of claims 1 to 8.

Description

Urban mobile network traffic generation method and device, electronic equipment and storage medium Technical Field The present invention relates to the field of communications technologies, and in particular, to a method and apparatus for generating urban mobile network traffic, an electronic device, and a storage medium. Background Urban mobile network traffic generation refers to the generation of mobile network traffic generated within the city. In cities, the number of mobile devices is increasing, including but not limited to smartphones and various internet of things devices. These devices access the mobile network through nearby base stations and generate diversified mobile network traffic while performing related functions. The mobile network flow reflects the activities of network domains in a plurality of areas and a plurality of time periods in the city, and has important significance for planning, constructing and optimizing the urban mobile network. However, the current urban mobile network traffic generation work mainly generates daily traffic change conditions, and does not fully consider the influence of urban environment, and cannot generate traffic with longer time scale and finer space granularity, so that the practicability and the accuracy are insufficient. Therefore, how to generate large-scale, real and reliable urban mobile network traffic becomes a current urgent problem to be solved. Disclosure of Invention The invention provides a method, a device, electronic equipment and a storage medium for generating urban mobile network traffic, which are used for solving the defects of low practicability and low accuracy of urban mobile network traffic generation in the prior art and realizing the generation of reliable urban mobile network traffic. The invention provides a city mobile network traffic generation method, which comprises the steps of extracting a plurality of entities based on city data information, constructing city knowledge patterns based on relationships among the entities, wherein the entities at least comprise base stations, areas, business circles, POIs, POI categories and brands, obtaining knowledge pattern embedded vectors corresponding to the base stations based on the city knowledge patterns, inputting the knowledge pattern embedded vectors and noise into a trained generator network for generating an countermeasure network model to obtain generated traffic data output by the generator network, wherein the generated traffic data is city mobile network traffic under the city data information background, the generated countermeasure network model comprises city mobile network traffic pattern generation countermeasure networks with different time scales and continuous time city mobile network traffic generation countermeasure networks, the city mobile network traffic pattern generation countermeasure networks with different time scales and the continuous time city mobile network traffic generation countermeasure networks are all based on knowledge pattern vectors of a plurality of base stations for transmitting known network traffic, obtaining generated traffic sample data, and training sample data of the base stations based on a discriminator network for transmitting the real traffic data and the generated traffic sample data, and the generated traffic sample data are obtained as real transmission data of the known traffic. The method for generating the urban mobile network traffic provided by the invention comprises the step of obtaining the knowledge-graph embedded vector corresponding to the base station based on the urban knowledge graph through a knowledge-graph embedded model. The urban mobile network traffic generation method comprises the steps of carrying out projection processing on real traffic data and generated traffic sample data to obtain real traffic data projection vectors and generated traffic sample data projection vectors respectively, carrying out processing on the real traffic data projection vectors and the generated traffic sample data projection vectors through a multi-layer perceptron, layer normalization and an activation function to obtain a discrimination result output by the discriminator network, and carrying out countermeasure training on the urban mobile network traffic mode generation countermeasure network with different time scales based on the discrimination result and a loss function. According to the urban mobile network flow generation method provided by the invention, the loss function is determined by adopting the following formula: D (S) represents a discrimination result of the real flow data; representing the discrimination result of the generated flow sample data; Representing a distribution of the real flow data; representing a distribution of the generated traffic sample data; representing uniform sampling between true-generated sample pairs; Representing the discrimination result of the sample; λ represents penalty